The Optimization of TSK Regression Model Based on Error Patch Learning Algorithm

Takagi-Sugeno-Kang (TSK) fuzzy systems are widely used in data processing due to their high interpretability. Patch learning (PL) algorithm is a new ensemble learning method and has attracted extensive attention, but it uses trapezoidal membership function, which will make the gradient discontinuous during parameter optimization and affect the convergence of the algorithm. In order to overcome the above problem, an adaptive FCM-based error patch learning algorithm is proposed in this paper. In addition, the proposed algorithm solves the problem of manually setting the number of Fuzzy c means (FCM) clustering rules, which is often used in regression problems. Simulation experiments are carried out on 12 real regression datasets and nonlinear functions, and the performance indicators are verified in multiple dimensions, which proves the effectiveness of the method.